Sample size from big industrial runs and batchess

[donald reinhardt]

Dear Folks- I have no knowledge of what is an adequate number of samples to take for an analysis from a total of 20,000 bottles of 1 quart from two runs that are each of 10,000 in number. I want to get an idea of the quality of the product at a reasonable confidence level 95% so that the product can be verified as acceptable for the marketing and the consumer.
1. Is a randomized sampling of each lot in order?
2. What number of samples should be tested for validation?
3. What is the best software out there that is easy to use and learn with?

Many thanks for your help and ideas. Don

 

[JohnM]

Donald,

Welcome to the forum.

1. Unless you know of some underlying variables in each run that would impact quality, then yes, simple random sampling is fine, but you may want to read up on SPC (Statistical Process Control), which will provide a set of tools, namely Control Charts, that will help you assess the stability and control of the process. A good source is Statistical Quality Control by Montgomery.

2. That depends on what margin of error you can tolerate, what measure you're interested in (average, proportion, etc.), and an estimate of the historical variability. Here's a pretty good source:

http://www.itl.nist.gov/div898/handbook/ppc/section3/ppc333.htm

3. Minitab is a good one - IMHO, the best bang for the buck.